Archive for April, 2007
No More Undergrad!
April 22nd, 2007
I am an undergrad no more! I finished my last exam on Friday and came back home to Thornhill that night. I don’t feel any different from when I finished exams in my previous terms. I think the fact that my undergrad career is actually over hasn’t sunk in yet. Overall, I had a pretty enjoyable term. The courses that I took were pretty interesting. Well, most of them anyway. The two courses I enjoyed the most were SOC 101 (Intro to Sociology) and CS 486 (Artificial Intelligence).
Sociology allowed me more insight into what is happening with the Canadian society, as well as brief discussions on societies in the global scale. I never really understood why Ams liked sociology that much, but now I have come to appreciate this field of study as well. The turning point for me is when we looked at myths that are widespread in our society, but are in fact all false. For example, people tend to have the belief that homeless people are lazy and have only themselves to blame for their situation. However, this belief is generally not true. We are socialized to be very individualistic, and the capitalist society awards individual achievements. So when we see people with problems, or if we ourselves have problems, we have a tendency to blame it on the individuals — but in actuality the problem is most likely a social problem.
My favourite course this term is AI. I’ve always had an interesting, or fascination, with AI. I think the challenge of having machines performing complex tasks that are usually done by humans is very interesting. It was cool to see the different techniques that AI researchers use to solve different problems. I enjoy computer science not for its practical values, but I love the theories behind everything. If I had to choose between a situation where I am in a situation similar to Bill Gates of Microsoft or Steve Jobs of Apple, or a situation where I’m just a researcher (in computer science) getting paid 60K, I’d take the research position without hesitation. I’m not faulting anyone for wanting money, but personally I just want enough to get by. I find more satisfaction in doing work that I find interesting than getting paid a six figure salary working a job I’m not very interested in. I’ve always enjoyed problem solving and puzzles. The feeling I get when I complete a proof or solve a very hard problem/puzzle is one of the most rewarding feelings I get.
Anyway, I still waiting for all my marks to come in. I have a good feeling about this term, and I’ll see if my good feeling is justified or not. :P
Computer Go
April 19th, 2007
I found an utility online that wraps java apps with a Windows executable called Launch4j. I used it to make an executable for the Computer Go program I programmed for my AI (CS 486) project. I’m very interested to see how the program would fair against different people. Download the program using this direct link!
If you don’t know how to play go, here’s a website to get you started. The rules are very simple and it doesn’t take long to learn how to play. But it does take a while to get good at it.
My program does no tree searches at all, and has VERY limited pattern matching on the game board. When I say limited I mean less than 10 patterns exist in my program — compared to thousands in most go programs. As I mentioned previously, my program determines moves almost entirely using influence and evaluation functions. One thing I did use that help a lot is machine learning. For my demo I ran 30 simulations and had the computer try out different parameters for its evaluation function and pick out the best configuration it has encountered so far. When my prof played it it did manage to beat her, but it had some obvious flaws. After that I decided to run it an additional 100 times, and it made a HUGE difference. It’s still not super intelligent, but I do think it fairs pretty well. Of course, once you’ve played it for a few rounds you might notice some weaknesses that you can exploit… this is unavoidable in almost all AI game playing programs.
AI Computer Go Demo Went Well
April 16th, 2007
I just came back from the demo of my AI (CS 486) project for my prof. I’m really glad I was able to get my program to play go at a decent level. Go is one of the hardest problems in computer game-playing due to many factors. For a while I thought I might have been too ambitious when I proposed to create an agent that can play go. I made a lot of progress when I first started the project back in February, but then I realized the approach I was using was not very good. Then the following weeks after that I hadn’t had the time to work on my AI project very much, so it just sat there taunting me until I finally resumed work just under 2 weeks ago. In th end I found a combination of using an influence function for determining good moves and a linear evaluation function for determining the best move to work well for my very simple go playing agent.
My demo went well today, and my prof liked what I have done so I’m very happy about that. She also told me that if I’m considering grad school I should let her know. So now it seems fairly likely I will be going to grad school sometime in the very near future.
STAT 443 Final is NOT Easy
April 13th, 2007
I wrote my STAT 443 (forecasting) final exam last night… long story short, it didn’t go well. The whole exam was done on Maple TA, which is some online test application created by Maple. First of all, all the things I studied for were not worth a lot on the final. I thought the final would be roughly the same as all our assignments and the midterm — which was mainly identifying, estimating and verifying models for different datasets. BUT, the final exam had a lot of questions on theory, which I didn’t study for in great details. There were also a lot of true and false questions, and some of those I wasn’t sure about because I don’t think there was any mention of them in the lecture notes. For example, one question stated “the Dickey-Fuller test is always conclusive (T/F)”… I’ve used the DF test, but I don’t know if it’s ALWAYS conclusive, so I put true because I haven’t come across a case where I’ve ignored the result yet, not like the Shapiro-Wilk test for normality.
And with about an hour left on the exam, the computers started to reboot. Some stupid Windows update thing forced everyone to reboot. And it took about 30 mins for everyone to reboot and login again. The server was being flood by all the STAT 443 students and couldn’t handle the load. So the exam was extended by 30 mins…. not like it really helped anyway, I was pretty much done by the time the computers started rebooting, but I stayed a while longer to try and figure out some of the questions I didn’t get. All in all, it was a pretty crappy way to kick off the finals.
Classes Over for Undergrad!
April 6th, 2007
I finished all my classes on Tuesday, which means I’m done classes for my undergrad career. It’s a bittersweet feeling. In a way I won’t really miss all the late-nights and all-nighters, but I will miss school. Being in school I’ve been kept sheltered from the real world — with the exception of co-op terms. It’s going to be very different to step into the work environment for a much longer period of time. In the academic environment you’re able to try out so many new things, but in the work environment there are so many constraints on what you can do. Not that I don’t enjoy work — there are many things you can learn at work as well — but I don’t think it’s as exciting as things you do in school.
Anyway, I got a call back from Workbrain, and they want me to start work with them in June. However, because they recently got acquired by Infor, they cannot officially give me an offer until the whole acquisition process is done. Hopefully everything will go smoothly and I’ll start work in two months so I don’t have to do any more job hunting… I hate job hunting.
The next few weeks for me will just be studying, and finishing my AI project. It’s going to be pretty busy, but it’s nothing I haven’t been through before. =P
